Meta Defends Muse AI Amid Allegations of Unauthorized Data Access

Meta VP of Communications Andy Stone and Superintelligence Labs executive David Singleton have both refuted a report by Inc. columnist Jason Aten, who alleged the AI read his private messages despite Full Disk Access being disabled. According to Singleton, the Mac version of Muse relies on a rigorous, three-stage permission protocol integrated directly into the macOS system interface. He noted that even if a bug existed within the application, the underlying operating system protections prevent unauthorized access to message data.

Singleton explained that users must manually navigate macOS Settings to grant permission, a process that triggers a mandatory app restart. He dismissed the AI’s own explanation—that it was syncing device notifications—as a "hallucination" or incorrect output by the model. Despite these technical denials, skepticism remains high. The company recently faced a jury verdict in New Mexico confirming it misled users regarding data practices linked to the Cambridge Analytica scandal. While Meta currently holds a top spot on the App Store, the incident highlights a broader trust deficit. Adding to the friction, YouTuber Matt Robb recently reported a separate issue where Muse allegedly mishandled a Facebook Marketplace transaction, leading to the unauthorized disclosure of his home address. Meta has acknowledged this specific claim and is currently investigating the potential error.